Fork me on GitHub
#onyx
<
2016-11-11
>
martinklepsch14:11:22

@colinhicks where's the current onyx-blueprint thing running? Might try to contribute something

martinklepsch15:11:34

I remember seeing it running somewhere though? Only GH Repo?

colinhicks15:11:32

It's not up publicly at the moment. I can put a new preview up later today if it would be helpful, @martinklepsch

michaeldrogalis15:11:27

Ah, yeah I was looking for the live copy a few days ago too. Thanks for the explanation @colinhicks. I’ll take a stab at doing a contribution on Sunday.

martinklepsch15:11:49

@colinhicks: ah ok, then im confusing it with this other thing I saw, sorry haha

colinhicks15:11:48

Sorry for the confusion. I took it down in part to simplify things as other folks start to contribute

colinhicks15:11:47

I'm AFK now, but will be back later to provide more details

Drew Verlee20:11:11

Any ideas on what would roughly need to be done for a Hbase Plugin? At a high level i would assume I would need to build something to provide an “exactly once” contract between onyx and HBase.

michaeldrogalis20:11:59

@drewverlee For a reader or writer? Reader needs a way to partition the data set into discrete chunks that can be read in parallel. Plugins do not need to provide any exactly-once mechanism. Onyx itself takes care of that completely.

Drew Verlee21:11:30

@michaeldrogalis I was semi interested in what it would take to do either. I was trying to compare it from what im reading about Apache Flink: https://ci.apache.org/projects/flink/flink-docs-master/setup/fault_tolerance.html, which discusses that the sources and sinks influence the guarantees your system has.

michaeldrogalis21:11:26

@drewverlee We’re getting close to having an alpha of the same fault tolerance system. Might want to wait a bit.